Tutorial: Building Your First AI Agent

2.3: Giving Your Agent Tools

Your AI agent can now answer questions and reference your knowledge base. But what if it could DO things—search databases, send emails, create tickets? That's what tools are for!

What Are Tools?

Think of it like this: Your AI agent is like a smart assistant with hands. Tools are what lets it actually take actions on your behalf.

💡 What Tools Can Do

Tools let your agent:

🔍 Search databases - Find customer info, order status, product details
📧 Send emails - Notify team members, follow up with customers
📅 Create appointments - Schedule meetings automatically
🎫 Create tickets - Log support requests in your system
🔗 Call APIs - Connect to any external service

Without tools, your agent can only TALK. With tools, it can ACT.

Example 1: Search Contacts Tool

Scenario: A customer asks "Do you have my email address on file?"

Your agent needs to search your contacts database to answer. Let's set that up!

Step 1: Create a Contacts Table

First, we need somewhere to store contacts. Let's create a second table:

Click "+ Create New Table"
Name it: "Contacts"
Description: "Customer contact information"
Click "Create Table"

Add some columns:

name (Text) - Customer name
email (Text) - Email address
phone (Text) - Phone number
company (Text) - Company name

Add a few test contacts:

Name	Email	Phone	Company
John Smith	john@example.com	555-0100	Acme Corp
Jane Doe	jane@techco.com	555-0101	TechCo
Bob Wilson	bob@startup.io	555-0102	Startup Inc

Name	Email	Phone	Company
John Smith	john@example.com	555-0100	Acme Corp
Jane Doe	jane@techco.com	555-0101	TechCo
Bob Wilson	bob@startup.io	555-0102	Startup Inc

💡 💡 Real-World Use

In production, this table would be populated from your CRM, database, or imported from a CSV file. For now, manual entries work great for testing!

Step 2: Give Your Agent Access

Now let's give your AI agent the ability to search this table:

Go back to your "Customer Conversations" table
Click on your "AI Answer" column header → "Configure"
Scroll to "Tools" section
Click "+ Add Tool"
Select "Search Table"

Configure the tool:

Tool Name: search_contacts
Table to Search: Select "Contacts"
Columns to Search: Select name, email, company
Columns to Return: Select all columns

Configure Tool: Search Table

Set up how your AI searches data

Tool Name

A clear name your AI will use to call this tool

🔍 Table to Search

Your AI will search the "Contacts" table when it needs to look up information.

Columns to Search

name

company

AI will search these columns to find matching records

Columns to Return

name

phone

company

AI will receive these fields from matching records

Tool configured! 🎉

Your AI can now search the Contacts table using the search_contacts tool.

Update your AI instructions:

Add this to your AI's instructions:

When a customer asks if you have their information on file, 
use the search_contacts tool to look them up by email or name.

Step 3: Test It!

Let's see it in action:

Go to your "Customer Conversations" table
Click "+ Add Row"
In the "Customer Question" field, type:
"Do you have john@example.com on file?"

Watch what happens:

AI Agent Using Tool in Real-Time

Watch the tool execution flow

Customer Question: "Do you have john@example.com on file?"

Status Updates:

🤔

AI reads the question

✓

🔍

AI decides to use the search_contacts tool

...

⚡

Tool searches the Contacts table

...

✅

Tool returns: "John Smith, john@example.com, 555-0100, Acme Corp"

...

💬

AI generates response

...

🎉 Your agent just used a tool! It searched your database and provided accurate information.

💡 What Just Happened?

Your AI agent:

Understood the question required looking up data
Chose the right tool (search_contacts)
Executed the search automatically
Used the results to craft an informed response

This is way more powerful than just answering from knowledge!

Example 2: Send Email Tool

Scenario: A customer requests a quote. Your agent should notify the sales team automatically.

Step 1: Add an Email Column

Let's add a column that can send emails:

In your "Customer Conversations" table
Click "+ Add Column"
Choose "Send Email" (find it under "Talk to Customers")
Name it: notify_sales
Label it: "Notify Sales Team"

Configure the email:

To: sales@yourcompany.com
Subject: "New Quote Request from [Customer Name]"

Body:

A customer has requested a quote:

Question: [Their original message]
AI Response: [What the AI said]

Please follow up promptly!

The system automatically fills in the customer's name, their question, and the AI's response!

Configure Send Email Column

Set up automated email notifications

Wait for these columns to be filled first

AI Answer(ai_response)

Email sends after the AI has generated a response

📧 Email Settings

Recipient email address

Subject

Use [placeholders] for dynamic content

BodyEmail content with [placeholders] for dynamic fields

Dynamic Placeholders Available:

• [Customer Name] - Customer's name
• [Their original message] - The customer question
• [What the AI said] - The AI's response

Preview

To:sales@yourcompany.com

Subject:New Quote Request from [Customer Name]

A customer has requested a quote: Question: [Their original message] AI Response: [What the AI said] Please follow up promptly!

Email configured! 📧

Emails will be sent automatically to sales@yourcompany.com when the AI generates a response.

Set Dependency:

This column should run AFTER ai_response is filled
This ensures the AI has already responded before the email is sent

Step 2: Configure AI to Use the Tool

Now teach your AI when to send emails:

Edit your "AI Answer" column configuration
Update the instructions:

You are a helpful customer service assistant.

When a customer asks for a quote or pricing information:
1. Provide a helpful initial response
2. Let them know the sales team will follow up shortly

Note: The system will automatically notify the sales team.

Step 3: Test the Complete Flow

Let's try it:

Click "+ Add Row"
Type in "Customer Question":
"I'd like a quote for your enterprise plan with 100 users"

Watch the magic:

Multi-Step Automation in Action

Watch the complete flow execute automatically

Customer Question: "I'd like a quote for your enterprise plan with 100 users"

AI Response Column

⏳ Waiting

Notify Sales Column

⏳ Waiting

Waiting for AI response to complete...

💡 🎯 Full Automation Achieved!

Now you have a complete automated flow:

Customer asks for quote
AI responds helpfully
Sales team gets notified automatically
No manual work required!

This scales infinitely—handle 1 quote or 1,000 quotes with zero extra effort.

Common Tools You Can Add

🔍

Search & Lookup

Search Tables

Find data in other tables

Lookup Order

Check order status

Find Customer

Retrieve customer info

Check Inventory

See product availability

📧

Communication

Send Email

Notify team or customers

Send SMS

Text message notifications

Post to Slack

Alert team channels

Create Notification

In-app alerts

📝

Data Creation

Create Row

Add data to tables

Create Ticket

Log support requests

Create Task

Add to to-do lists

Schedule Appointment

Book calendar slots

🔗

External Services

HTTP Request

Call any API

Run Custom Code

Execute JavaScript

Query Database

Direct database access

Call Webhook

Trigger external systems

Mix and match tools to create powerful workflows!

Combine multiple tools to automate complex processes. For example: search customer data, draft a personalized email, and send it automatically.

Best Practices for Tools

1. Give Tools Clear Names

Bad: tool1, helper, function
Good: search_contacts, notify_sales, create_support_ticket

Clear names help the AI understand when to use each tool.

2. Write Specific Instructions

Tell your AI WHEN to use tools:

❌ "Use the search tool when needed"
✅ "When a customer asks about their order, use search_orders 
   with their email address to find their order status"

3. Test Each Tool Separately

Before combining tools, test them individually:

Test search tool works correctly
Test email tool sends properly
THEN combine them in workflows

4. Set Up Error Handling

What if a tool fails? Configure fallback responses:

If search_contacts finds no results:
"I don't see that email in our system yet. Would you like to provide 
your contact information so we can add you?"

Real-World Tool Combinations

Customer Support Agent

search_tickets - Find past support issues
search_orders - Look up order status
create_ticket - Log new issues
notify_support_team - Alert team for urgent cases

Sales Assistant

search_contacts - Find customer records
check_inventory - Verify product availability
calculate_quote - Price calculation tool
notify_sales - Send leads to sales team

Order Processing

validate_address - Check shipping address
check_stock - Verify inventory
create_shipment - Generate shipping label
send_confirmation - Email customer

Understanding Tool Execution

Here's what happens when your AI uses a tool:

Tool Execution Flow

What happens when your AI uses a tool

🤔

AI reads the message

Understands what the customer needs

🧠

AI decides which tool to use

Matches the need to available tools

"They want to know if we have their email → use search_contacts"

🔧

AI prepares tool parameters

Extracts relevant data from the message

Pulls "john@example.com" from the question

⚡

Tool executes

Searches the Contacts table

Returns results

💬

AI uses results

Incorporates the data into its response

"Yes, I found John Smith with that email..."

This all happens automatically in seconds!

This all happens automatically in seconds!

When to Use Tools vs. Knowledge

Use Knowledge Base for:

✅ Answering policy questions
✅ Explaining how things work
✅ Providing instructions
✅ Sharing information from documents

Use Tools for:

✅ Looking up specific customer data
✅ Checking real-time status
✅ Taking actions (sending emails, creating records)
✅ Searching structured data

Use Both Together: Your most powerful agents use knowledge for context and tools for actions!

What You've Learned

By completing this section, you now know how to:

✅ Add tools to your AI agent
✅ Configure search tools to find data
✅ Set up email tools for notifications
✅ Combine tools for complete workflows
✅ Write instructions that guide tool usage
✅ Test and verify tool functionality

Your agent went from chatbot → action-taking assistant!

Try This Challenge

Challenge: Create a support ticket tool

Create a "Support Tickets" table
Add a "Create Ticket" column in your conversations table
Configure your AI to create a ticket when:
- Customer reports a problem
- Issue can't be resolved immediately
Test it with: "I'm having trouble logging into my account"

Bonus: Add a notify_support_team email that sends when high-priority tickets are created!

Ready to make your agent's personality shine? Let's customize how it communicates!

Previous: Adding Knowledge
Continue to Section: Customizing Personality →

Tutorial: Building Your First AI Agent ​

2.3: Giving Your Agent Tools ​

What Are Tools? ​

💡 What Tools Can Do

Example 1: Search Contacts Tool ​

Step 1: Create a Contacts Table ​

💡 💡 Real-World Use

Step 2: Give Your Agent Access ​

Configure Tool: Search Table

Step 3: Test It! ​